The present work advances the research on hedge fund returns in three main areas. Firstly, their statistical properties are assessed in order to understand by what degree the returns of this alternative asset class are subject to non-normality, autocorrelation and heteroscedasticity. Secondly, state-of-the-art econometric approaches are used for the purpose of analyzing whether and to what extent monthly hedge fund returns are forecastable. Thirdly, an effort is made to identify and explain which economic risks affect the performance of the different hedge fund strategy styles in which way. The empirical results suggest that monthly hedge fund returns are forecastable by means of multivariate regression models which rely on economic predictors such as changes in interest rates or changes in business outlooks. Accounting for the fact that hedge fund returns are non-normally distributed, heteroscedastic and time-varying in their exposure to pervasive risk factors, the devised econometric models are found to deliver significant out-of-sample predictive power. The thesis at hand also documents that the interdependencies between the monthly changes of envisaged risk factors and the subsequent hedge fund returns remain remarkably stable throughout time. In essence, the performance of hedge funds appears to be sensitive to common business cycle movements. Altogether, the results are relevant to researchers in search of a description and application of contemporary return prediction methods as well as to investors in need of a better understanding of the drivers of hedge fund returns.